Hey, thanks for the report. This is a known issue. Context summarization can miss important details and cause the AI to go off track, especially in long chats with big context windows.
A couple of things that might help:
Start a new chat more often, before the context window fills up and summarization kicks in. You can use @ references to quickly pull in the right files for the new chat.
Use a .cursorrules file or a project rules file with key instructions. This helps the AI keep critical context even after summarization.
To help us look into your case, could you share the Request ID? You can get it from the chat context menu in the top right, then Copy Request ID. If you can, grab it from the message where the response went off track.
The team is aware of summarization quality issues. Your report helps increase visibility, especially for Team plan users.
